Optimizing LEDs / the dilemma of tall enclosures / tank build log (possibly pic heavy)

Calvin, it looks like a very nice enclosure! It looks pretty professional looking.
A couple comments.

1. Because of the inverse square law, it is pretty tough to get usable space from tall enclosure.

From physics of light, if you mount the light source far away, then the difference in the intensity between the top and the bottom of the enclosure is minimized. But it is not practical (you need high ceiling and need narrow beam optics).

If you can mount the light to the front door (maybe 2 feet from the bottom), then you can have more usable area at the bottom. Or with your image C, you can use narrow beam pattern to target the lighting of the bottom area.

2. You might want to consider COB-LED for the next addition. It is much easier (maybe < 20 min to make a unit). I didn't have enough DIY skill, so I didn't try DIY LED until COB becomes cheaper. Cree CXA3070 ($40) or Bridgelux Vero 29 ($30) driven at 50W is pretty good. If you want 30W-class, then Vero 18 is much cheaper (around $13-15). I've been using Cree CXA3070, but the lower price of Vero is attractive, so the next addition (I'm ordering them next week) is Vero 29 (which can be driven at 70-100W easily with a small amount of efficiency penalty). The driver for 50W or less is around $13, but above that the driver become expensive (around $30 for cheap 80W or $40-50 for name-brand 80-100W). So 50W is the sweet spot now. I used to think warmer white (3000K) is nicer, but I'm going for the higher absolute light output of 4000K or 5000K next.

3. Passive cooling is possible, but you need to under drive the LEDs and need fairly big and expensive heatsinks. With active cooling, $10 CPU cooler is enough. Also if you compare 30W active cooled vs 30W passive cooled, active cooled one is "usually" more efficient (more light per watt) even after you consider that the fan consumes extra energy (cooler LED emits more light). So I would recommend going with active cooling. If the noise is the issue, you can try expensive quiet fans.

4. If you want to optimize the efficiency, did you consider attaching the mylar reflective sheets to the enclosure walls?

I'm looking forward to seeing how your project progress! It is surprising that there aren't so many orchid people doing DIY LED yet (well, digging through the information at first was pretty tough for me).

 

I shouldn't even mention it because I don't know where you can get them, but I have seen LED's mounted to a 24" aluminum heat sink/backer.
This was just a few months ago at a lighting supplier. These were brighter than a T5 6400k and used less than half the juice; not as hot as T5!
If anyone knows where you can get them for DIY lights... Let me know!

 

Don't know if this viable for you, but one can get a string of outdoor LED's for outdoor(with the gel coating) from China. For under $30US, on can get 16ft of this stuff. If you can find a way to hang it low or just put it on the bottom it might solve your problem!?

Picture: 16ft of super bright LEDS attached to piece of aluminum screwed in turn to old T5 fixture.

 

Forgot pic:

 

Natureman, are you talking about the LED Strip light? When I looked into it, they were pretty inefficient. Check message #14 in the following forum:

http://www.candlepowerforums.com/vb/showthread.php?365587-Recent-LED-Strip-Comparisons

If you are looking for strip-like heatsink for DIY, this place have it:
http://www.heatsinkusa.com

If you want to make a long fixture, Bridgelux Vero 10 based project may be the best bang for the buck.

Some people use 3/4 to 1" aluminum U-channel from HomeDepot as the cheap heat sink (I'm not sure if this is sufficient for Vero 10, though).

 

Be careful about using such lights. They're intended to provide humans with light, not plants.

White LEDS are very bright because they emit predominantly in the green part of the visible spectrum that the human eye is most sensitive to, not the red and blue ends that plants are.

I'm not saying they cannot be used for plants, but you're going to want extra wattage to compensate for the spectral imbalance.

 

No, I'm not talking 3528 or 5050(that's what I'm showing in my picture, though).
I'm talking about a brand new 'hard' strip; the LED's are integral to the aluminum plate and they're really bright! Brighter than a 6400 T5. I saw it in a 24" strip. So bright you can't look straight at it....
Really, my point is that new LEDS are coming soon(if not already here) and the smart money's on holding off until we see this new stuff. I find it very exciting!
The lighting company I mentioned, had one while I was there buying some repair parts. After I explained I was into orchids, they ordered enough to make a 4' light comparable to a 4' 2 bulb T5... They had me back to show me and it was very impressive... They never called me back with a price, though?

 

Natureman, I wonder what this new stuff you are talking about. It sounds interesting. If you can get more info about this (e.g. who makes it), I'm interested!

Calvin, you are right; spotlight effect is the disadvantage of big COBs (the lower leaves get shadowed easily)

My setting is like this. 4x4x8' grow tent with two levels (the top floor is about 4' above the ground). The lower level has 4x 50W COB (but I'm running only 3 at this moment). Since they are inside of a reflective grow tent, the spotlight effect is somewhat buffered. No optics or reflectors are used for each, so it has a wide beam angle. In other words, there are some lights from more distant LEDs reaches to the lower leaves. But I admit that, compared with florescent, there are more shadows. Since you have the skill, bunch of smaller COBs could be better. But I have been focusing on larger COBs, so I don't know which small ones has the best price/performance. I know non-COB Cree XML-2 is pretty good, too.

My LEDs are inside of the enclosure, and the humidity usually reaches to >90% at night, I don't know how this will influence the longevity of LEDs. I have accidentally splashed water to LED surface a couple times, and it's probably not a good thing. Hopefully, mine doesn't break prematurely... When you look at the data sheet, both bridgelux and Cree mentions that they didn't have failure at 85%RH.

Waterproofed enclosure would be nice, but I wonder how it will influence the heat management; higher temp means lower efficiency. I haven't investigated about waterproofing, but this one seems to be interesting: LED light engine is sealed for outdoors - Bridgelux. It has Vero 10 with water-resistant enclosure. But it is probably pricey.

When I put a 30W COB inside of 2x2x3' enclosure, the enclosure became too warm. So, this could be an issue if you put too much LEDs inside.

Vero 10 is about 2cm diameter. What's the dimension of the triangle heatsink? But you probably need to under-drive quite a bit to use passive cooling with this heatsink (just a guess).

You are right there are quite a bit of wires (4x driver and 4x 12V for the cooling fan), I just use twist tie etc. in an attempt to organize. I'm using a single 12V AC/DC adapter to drive the four cooling fan, so this does reduce the wire a little bit. I want to keep the enclosure cool, so I have the drivers outside of the enclosure. Then 4x CC drivers + 12V are plugged into a power strip with a switch. It's not pretty, but it works ok for me.

If you use relatively expensive driver with high voltage, you can connect a few in serial (similar to your XP-G) to reduce the wire. Forward voltage of Vero10 is around 26V for Vero 10, 30V for Vero 13 & 18 , and 40V for Vero 29 and Cree CXA3070. Meanwhile HLG-C series is good for this purpose because it can supply high voltage needed for serial connection.

http://www.meanwell.com/webnet/search/seriessearch.html
http://www.meanwell.com/webapp/product/search.aspx?prod=hlg-120h-C
HLG-120H-C1400 can drive 2x CXA3070/Vero 29 to give 100W total
HLG-120H-C1050 can drive 3x5 Vero 18 to give 90-150W total
According to octopart.com (price look-up engine), it costs around $60.

For V10 with lower current (350mA), you can find much cheaper drivers, e.g. this can drive about 4 Vero 10 to give about 35W.

 

Ray, we have talked about this (how much light is lost due to the phosphor of white LEDs vs blue and red monochromatic LEDs). I realized that the figure 4 of this paper has some relevant info:

http://www.plosone.org/article/info:doi/10.1371/journal.pone.0099010

I would say that cool white isn't so bad (about 20% lower output than blue). This figure doesn't give the complete answer because it doesn't consider that red light is slightly better than blue light for photosynthesis. To consider this point (the wasting energy for green light which you are talking about), we can make a rough calculation with YPF/PPF values from the table 1 of this poster. YPF/PPF is 0.86, 0.7, 0.96 for cool white, blue, and red, respectively. Now if we assign the photon output of Blue as 100%, cool white is about 80%, and red is about 95% (eyeballing from PlosOne figure). So YPF (which is a measurement which incorporate the inefficiency of green light for photosynthesis) would be 0.86 * 0.8 = 0.688 for cool white, 0.7 * 1.0 = 0.7 for blue, and 0.96 * 0.95 = 0.912 for red. So red is the best for photosynthesis, and cool white is about 75% in terms of YPF. I would say that it isn't so bad. Also, I'm not sure how much money company is putting into improving the red LEDs. So if we consider the COB advantage, white LEDs may not be so bad at this moment.

But there are some issues with this kind of simplistic calculation as you imply (effect of quality of light on biology of plants). But the spectra of LED seems pretty decent compared to the spectra of most florescent light.

 

Naoki -

I have the same philosofy as you about growth light. And I have used white leds (with good spectrum) for 4-5 years for most of my collection. And I have good results. Just as good and even better than with "extra-fancy-expensive-science-growth-light".

My good friend Magnus A (on this forum) who is a ph.D. in photosynthesis backs up the same thoughts about white light with a good spectrum as a good alternative.

------------------------------------------------------------------------------------------------------------

To return to the OP's original question I would recommend using PAR38 lamps with 60° lenses to focus light down on the lower parts of your cabinet. This is what they look like:



They can be found in different wattage. 12,15,18W. The one on the picture is 18W with mixed Kelvin diodes (70% 6500K and 30%4500K) The neat thing about these are that you use a standard lamp socket for them and they are not that expensive. I even use them in my greenhouse to reach areas down at he floor.

_________________________________________________________________________________________

And when speaking about LED-strip lights.

Latest technology LED with a crazy light output of 150 lumen/W. 5 mm wide, 3 mm thick. The stick on the picture is 60 cm, 20W and 3000 lumen.
Mixed diodes of 30% 4500K and 70% 6500K.

 

Do you have a pick of that LED stick off? Is it pre wired to the receptacle?

 

Pretty cool, Mikael! Is it a product you are developing or is it available from some company? If the former is the case, you might not want to disclose too much details, but I'm curious what emitters it is using (and driving current). I guess you'll probably need 40x Cree XB-D or XT-E etc driven at pretty low current to achieve 150lm/W (and to allow passive cooling). It seems to be a pretty luxurious design (for low-volume DIY people, it costs at least $40 for the LEDs to get 20W)! One can get similar efficiency (and total of 20W) with 6x XM-L2 or 10x XP-G2 (both would go around $30), but they probably end up with a bigger stick (not as slick looking as yours).

As a comparison, with Cree CXA3070, you need to get relatively-hard-to-get high-efficiency bin (AB, around $40-50) to achieve that efficiency (after production, they test the LEDs and categorize them into different efficiency "bins", so the same model, e.g. CXA3070 5000K can be available in several flavors of efficiency with slightly different price). Then you have to drive it at a low current (34W). Efficiency (lumen/W, or PAR/W is more relevant to us) increases with lower driving current for any LEDs. But we have to make a compromise, because you need lots of LEDs (which become costly) if you are driving each LED at a low current. This is a part of reason why efficient grow light is more expensive. After setting the lumen/W as a constant (150lm/W), you can get a bit more total lumen per dollar (or euro) with COB. But the strip (or panel) form factor is nicer than the point light source of COB!

For people who like to waste time in comparing LED data (like me), this cree site is pretty cool:
http://pct.cree.com/dt/index.html

I didn't know Magnus's field was in photosynthetic research! It is a very cool field. It is amazing how much (and how little) we know about it. I would expect that companies will be pouring more money to plant optimized LEDs (we have to deal with our growing population size), but the progress seems to be slower than white LEDs.

 

I was just finishing my order of Vero 29, and I noticed this relatively new LEDs.

http://www.ledsmagazine.com/article...es-board-level-modular-led-light-engines.html

I haven't checked the details, but I wonder if this is something similar to what Mikael is showing, and what Natureman is talking about. Luxeon XF-3535L seems to achieve 140lm/W to 157lm/W (3000-5000K).

For example, 4000K, 16.8 W, 525x5.25mm gives 2464lm (147lm/W), and costs $19.37:
http://www.digikey.com/product-detail/en/L235-4080AMLF5WAD0/1416-1355-1-ND/4966152
The performance/cost seems to be pretty comparable to COBs from a quick calculation.

Calvin, this might fit to the triangle heatsink you found (although I don't know the heatsink requirement for this strip light).

 

First, the easier Q to answer. For our purpose (grow plants), you should definitely go with the lowest CRI for LEDs (for a given color temp). To achieve higher CRI (flatter emission spectrum), they need to use "more" phosphor (well, this may not an accurate description), so it decreases the efficiency. As you have noticed higher CRI LED has a lower lumen. Similarly, people has calculated PAR from LEDs with different CRI, and lowest CRI for a given color gives the best PAR (almost always). I don't know the answer to the questions about CCT in your 4th&5th paragraph. But simply going with the lowest CRI for a given K is best for us.

You mentioned phosphors cutting out the green. It is probably similar to what's going on, but I don't think it is a quite accurate description, neither. Phosphors get excited by blue light, and emit reddish light. Somehow they are making the wavelength range of this emitted light wider (I don't know the exact mechanism, here).
The following doc has a figure to show this (the top figure of p.7):
Cree XLamp CXA3070 LED Data Sheet - Cree, Inc.
3000K high CRI has a beautiful, flat emission. 5000K comes in CRI of 70,80,90, but 3000K comes in a bit higher CRI of 80 and 93. I guess 5000K is dominated by blue peaks, so it is more difficult to create higher CRI.

Choosing the color temp is a bit more complicated. You get more PAR from cool white than warm white. People who grow photoperiodic plants want to have fair amount of red (for flowering), so they prefer 3000K. But many (not all) orchids are NOT photoperiodic. I have both cool and warm white COB, and my orchids seem to grow/flower OK in either one. So similar to Mikael, I went with 5000K and 4000K for the new order to get more absolute PAR (I'll measure and compare the actual PPFD and report it back eventually). Also in several plants, bluer light contributes to more compact plants (the 2nd link in message #15 contains this info), which may or may not be great for orchids.

 

I decided not to use reflectors. When I was building, I considered getting one of ledil reflector. Here is a link to Vero 29 compatible product:
http://www.ledil.com/products/?search=true&led=VERO29
You'll need to calculate what beam angle you'll need. I think you'll likely getting the LEDs from either mouser or digikey, and the last time I checked they didn't have led reflectors in stock. In this case, you can build one without reflector, and see if you need to order a reflector later.

Learning about phytochrome-mediated photoperiodism is pretty interesting! Another less-understood light sensor is cryptochromes (blue receptors), which influence the shape of plants. I looked up Fundamentals of Orchid Biology by Arditti, and Table 5-29 lists factors that control flowering for a fair number of species(3.5 page, double column table). I used to think most orchids do not show photoperiodism (this seems to be true for most Dens, Phals, Paphs), but there are more photoperiodic orchids than I expected. So there may be a benefit of going to 3000K or 4000K instead of 5000K.

Yes, it's a bit difficult to give lots of seasonal variation here, but there are about 10F difference between winter and summer. Plants seem to sense the seasonal change (even though the natural light is completely cut-off by the grow tent). But there are some plants which doesn't seem to flower (Phal. wilsonii and its relatives), so I'm experimenting with a larger seasonal changes for some species.

I forgot to comment on your link to a heatsink. I didn't see the dimension of the area where COB attatches, do you know if it has an appropriate size? Many people use this for Vero 29/CXA3070 class:
http://www.amazon.com/ARCTIC-Alpine-11-Plus-Cooler/dp/B00606OHQQ/ref=sr_1_1?s=pc&ie=UTF8&qid=1411238141&sr=1-1&keywords=arctic alpine 11
At this moment the price is pretty high ($17.69), but usually the price is $10. You can set the price tracking option and let camelcamelcamel to email you when the price goes back to the normal price. If you can't wait, this place has it for $12 + some shipping. I use old-left over heatsinks (mostly from old computers, but one is from a Kenmore washing machine).

Here is a link of a pretty good DIY COB project.